Qualities of good seminar talks

28Oct07

Update: included a few more points on 11 Nov 2007.

A big part of doing science involves `taking your show on the road’ and giving seminar talks. Seminars are a way to promote your current project. More importantly, they are a way promote yourself to other physics departments that might later be interested in hiring you. Based on very limited experience, here are my obsevations about what makes a good science seminar/conference talk sparkle.

Never go over time. This is a cardinal sin. For every one person in the audience who is interested in the finer details of your work, there are maybe five who just want to get back to their own work. If you go over time, those people will hate you. Factor in questions (during and after your talk) when you practice timing your talk.

Seriously, NEVER EVER go over time. This is really important. A good speaker can control his or her own presentation, even if that means cutting some sections out to stay within time. Know ahead of times which sections are expendable. Keep track of the time using the clock in the room, the session chairperson, or your own watch. Oh, and try to avoid making it obvious when you’re checking your watch. If you check your watch, then everyone is going to check their watches. Include slide numbers or an outline; people are more likely to stay with you if they know you’re almost done.

Contingency plans. Don’t assume compatibility! Instead of uploading your talk as a .ppt, use pdf. Do you have a back-up on a USB memory stick or a copy on your own webspace in case the file upload doesn’t work? If you’re using your own laptop, do you have your own set of AV cables to connect to the projector? In the worst case, do you have transparencies for an overhead projector? It might be a bit overkill… but this talk might be the difference-maker for that postdoc position you want. No pressure. Include your name on every slide. Be sure to do this for conferences and workshops, where your audience is likely to forget your name after your introduction. Your title slide should have your name and institution so that people know how to look you up. (You have a department webpage with contact information, right?) Having your name handy also helps when people in the audience want to interrupt you to ask a question.

Important ideas go first. Good seminar speakers realize that the audience’s interest peaks early in the talk. Put the `big idea’ of your talk in the beginning, and then flesh out the ideas with the rest of your time. This is also nice because in case you run out of time, you can cut your presentation short without losing the point of your talk.

Frequently recall your logical flow. The main difference between a chalk board talk and a slide-based talk is that on the chalk board ideas stay up until one absolutely has to take them down. This means you can refer back to equations or an overall outline as you progress. In a slide-based talk one doesn’t have this luxury, so be kind to your audience and frequently remind them what equation (3c) of slide 5 was when you refer to it later on. (This is also why chalkboards are much better for lecturing courses!)

Speak with enthusiasm. Even if you’ve done something terrific, people will be tearing their hair out if you speak in a monotonous voice. Show that you’re excited about your work. Show confidence; stage fright will be interpreted as a lack of conviction in your own work.

Speak clearly. I was fortunate in high school to have participated in the academic decathlon, which involves giving prepared speeches. Our team would do voice exercises for enunciation, projection, and clarity of speech. I’m not sure if I’m any good as a speaker, but I know I’d be much worse if not for this tid-bit in my background. As a speaker, you have to speak at an appropriate volume and project your voice. (Note: these are different! A good speaker can send a whisper to the back of the room.) Never mumble your words — everything you have to say is important, even if it’s a side-comment — and practice good diction. This may require a bit more work if you’re shy or if English isn’t your primary language. Sorry, it’s not fair, but it will make a difference. Don’t be afraid to ask people in your department to help you practice.

Interact. Good lecturers interact with their classes, and to a certain extent a good seminar speaker can engage the audience without being too pedantic. Give credit to previous work by members of the audience, socialize during pre- and post- seminar refreshments, and treat questions as a dialogue. If the department that is hosting you offers to take you out for dinner, go out with them — but understand that they may be viewing this as an informal postdoc interview, so try not to get drunk, yeah?

Pace yourself. Just because you’re not getting quesitons, it doesn’t mean everyone understands you 100%. Understand that it takes some people more time to digest information. Pause after important points to highlight what you’ve done and to ask for questions.

Make it accessible. You must realize that not everyone in the audience is an expert in your subfield. Despite that, these of people (especially students) come to your talk because they’re interested in what you have to say. Thank them by making at least the beginning of your talk accessible to them; even if only the first few slides are accessible to a general audience, and even if your talk becomes rapidly more technical as you progress.

Label your graphs! Your graph needs, at the minimum: a title (what you’re plotting), a y-axis label, and an x-axis label. If you don’t have at least these three elements, then your graph sucks and is unintelligible. Even if you do have explicit labels, spend time discussing your graph instead of just showing it as proof that you’ve done numerical simulations. Explain qualitative features in terms of whatever model you’re presenting. Don’t hide behind graphs.

Cite any results, even your own. Often talk slides are available to the public. Figures and results should have citations, so that people can attribute them properly when they become interested in your work.

Have a `further reading’ slide. This is different from just including references (which you also must do!). If you give a good talk, people will be interested in reading more about what you’re doing. Give references to good review articles (starting at a level accessible to postgrads in your field) as an `extra slide’ in your talk. Most talk pdfs are now saved on conference servers so that attendees can access them in the future.

Indicate what should and should not be taken on faith. You can only fit so much into your talk. Sometimes you have to skip a few steps, that’s fine. But be sure to clearly indicate when you’re doing this. This way audience members won’t get confused when they can’t follow your steps anymore. Conversely, emphasize a logical progression if it is critical to your argument. If the `take home’ message of your talk depends on following a certain deductive process, make sure the audience understands how each step depends on the previous ones.

Synchronize the visial and aural. We’ve all grown up watching television. You must synchronize the visual and aural information you present. Don’t put up anything until you’re actually talking about it. Otherwise your audience will stop listening to you and start reading your slides. It’s better to have five different slides with digestible chunks of information rather than one slide with many paragraphs. We, the audience, are easily distracted.

No essays! On that note, your slides are not a substitute for a paper. Don’t be afraid to just use bullet points and sentence fragments. And you should almost never have full paragraphs on your slides. Remember: the slides are there to emphasize and compliment the things that you say, not to replace your voice. (This is the difference between slides and conference proceedings.)

Be a tease. Sometimes you absolutely must have everything on one slide. Examples include the outline of a complicated proof, a big circuit diagram, or any other presenation of how many ideas connect. Don’t put everything up at once! You’ll only confuse your audience. Put up pieces of the slide one at a time as you talk about them and gradually reveal the entire picture. This way the audiences focuses on each piece as you describe it.

Timing is important. Plan out when you say things relative to when you turn to the next slide. This is especially key if you’re telling a joke. (If you want a good example of how to tell jokes in a presentation, look up Demitri Martin’s comedy.)

Make it look nice. You don’t have to be a graphic designer to make a nice presentation, but use common sense. Try to keep a unified theme (e.g. don’t change backgrounds for no reason), use a foreground and background that contrast one another, use legible fonts and font-sizes. If you use a lot of equations, use an equation editor or beamer instead of trying to `hack’ it on powerpoint.

Arrows and diagrams. Break free of the mold that every slide is a series of bullet points. Include arrows and diagrams that emphasize logical flow or structure. It’s ok to be a little non-linear. Beamer users may want to consider usng beamer arrows.

Answer questions. Be sure to leave time for questions, and be ready for questions during your talk. If you have an idea what kinds of questions are likely, include `extra slides’ at the end of your talk so you can refer to them during the question session. (Tip: be sure not to include these slides in your slide numbering, or people will get scared and think you’ll be talking forever.)

Learn when to say `I don’t know.’ Try to answer every question. But don’t ever say something you’re not comfortable saying as a scientist, and certainly don’t burble on like the orchestra on a sinking Titanic. If you can’t answer a question, politely say so and move on.

Learn HOW to say `I don’t know.’ When you’re a beginning grad student, you can get away with answering questions with `I don’t know.’ But there will always be a question just around the corner that you can’t answer. Learn how good scientists respond to tough questions. Even if you don’t know the answer, offer some relevant information that you do know. Learn how to think on your feet to address (if not answer) these questions. Don’t worry, it will come with experience.

Be playful. This goes along with being enthusiastic about your work. You can take your work seriously without taking yourself seriously. Show the audience that you’re someone who they would like to work with. (But see note above regarding timing.) Be sure to keep it tasteful and professional, though.

Say `thank you.’ Don’t forget to thank the conference organizers or the department that is hosting you. And always acknowledge your collaborators.

Seminar Bingo. If you get stuck in a bad seminar, you might want to have a couple of these handy in your notebook: PhD Comics seminar bingo, physics seminar bingo.



One Response to “Qualities of good seminar talks”

  1. 1 Steffen

    Good posting, Flip. Cliff Burgess gave a talk in Cambridge today – he used a total of 114 slides…



%d bloggers like this: